Self-Improving Memory MCP

--- approved: - mcp__memory__create_entities - mcp__memory__create_relations - mcp__memory__add_observations - mcp__memory__search_nodes - TodoWrite --- # Manual Checkpoint Command Save current conversation state to memory before it gets too long or you need to pause work. ## Usage ``` /checkpoint [optional-description] ``` ## What this does This command triggers the Pre-Compact Interceptor Agent manually to: 1. **Capture complete session state** - Everything discussed and decided - All files modified - Pending tasks - Key context and decisions 2. **Save to memory** - Creates checkpoint entity in knowledge base - Links to all related work - Preserves TodoWrite state 3. **Generate continuation summary** - Compact summary to copy/paste - Resume instructions - All context preserved ## When to use - Before taking a break - When context is getting long (check token count) - Before starting risky/complex work - At natural breakpoints (milestone completed) - When you want to continue later in fresh conversation ## Example ``` /checkpoint before implementing authentication ``` ## After checkpoint You'll receive: 1. Confirmation that state is saved 2. A continuation summary to copy 3. Instructions to start fresh conversation ## To resume later In a NEW conversation, paste the continuation summary and Context Recovery Agent will automatically load the full context from memory. --- **Now activating Pre-Compact Interceptor Agent...** You are the Pre-Compact Interceptor Agent. The user has manually triggered a checkpoint. Follow your instructions in `.claude/agents/pre-compact-interceptor.md`: 1. Capture complete current session state 2. Create comprehensive checkpoint entity 3. Save to memory with all context 4. Generate continuation summary 5. Provide clear resume instructions Additional context from command: - Manual trigger: yes - User description: "{{ARGS}}" (if provided) - Treat this as high-priority checkpoint Begin checkpoint process now.